Geochain AI: Automated Spatial Analysisusing Large Language Models and Chain of Thought

Afzal. A1, Amarnath.S2, Mrs. S. Divya 3, Dr. S. Akila4,Dr. P. Dhivya5
Department of Computer Science and Engineering, Dr. M.G.R. Educational and Research Institute, Chennai
600095, India.

Abstract— Executing spatial analysis is important when determining where to build or provide services that are in a region due to trends and changes in the geography of some areas caused by various environmental factors. The standard techniques of estimating spatialrelationships often require desktop GIS that are manually used with complex GIS software packages like ArcGIS and QGIS. The typical use of desktop GIS has created a great deal of manual work by GIS professionals who are required to run each geoprocessing task in a series and perform many of these tasks manually using desktop GIS solutions. The way these workflows are produced and produced does not lend itself well to effective and efficient production. Therefore, to increase efficiency and effectiveness to execute any analysis for geospatial based queries for automated execution engineering/production against any geospatially-based question we have developed a framework to orchestrate the execution of geoprocessing tasks using LLMs and CoT reasoning. The geoprocessing tasks result from the LLM determining the subtasks necessary to complete the geospatially based request with the use of RAG modules. The RAG modules allow the LLM to secure clinical documentation that is relevant in order to execute on the identified subtasks (e.g., ArcPy, GDAL, and GeoPandas). Index Terms— Spatial Analysis, Large Language Models, Geographic Information Systems, Chain of Thought, Retrieval- Augmented Generation, Tool Orchestration, Location Intelligence.